Why it may be 'Back to the Drawing Board' for pollsters after this election
If Mitt Romney wins the election tomorrow (or in the next couple of weeks), it is likely that pollsters will be in for a period of soul searching not seen since they predicted Dewey over Truman. While they may be able to claim that some of the ones they didn't get were in the margin of error, the fact is, their polls have consistently given Obama a lead in states that it is very possible Mitt Romney might win tomorrow.
The popular notion of bias against the GOP or in favor of Democrats might fit some pollsters (PPP), but the problem pollsters will probably have to deal with has more to do with statistical analysis than partisanship.
Where did they go wrong - if they went wrong at all?
First, Jay Cost on Romney's thrust into Pennsylvania in the campaign's waning days and the disconnect of the polls:
The Romney campaign seems to have committed to a late push into Pennsylvania, to the derision of Team Obama. The latter sees this as a desperation ploy by a foundering campaign, similar to John McCain's late entrance into the Keystone State in 2008. Is that right?
I don't think so.
For starters, this is not 2008. The national polls are showing a tie at the moment, and as Sean Trende ably noted last week, a nationwide tie does not square with the kind of leads Obama "has" in the swing states. Romney can't be down in Ohio by 3, tied in Florida, Virginia, and Colorado, yet still be running neck and neck with the president nationwide. You literally cannot find the votes for the GOP, unless you assume that California is a toss-up and New York is set to go Republican for the first time in thirty years.
So why head into Pennsylvania, a state the GOP has not won in twenty years?
Liberal proponents of the "emerging Democratic majority" love to talk about shifting demographics in their favor, but they regularly ignore the many ways in which demographics have shifted toward the GOP. For instance, Nevada and its six electoral votes tipping red to blue is all the rage, but West Virginia and its five, Kentucky and its eight, Arkansas and its five, and Missouri and its ten switching blue to red is ignored.
And so it goes with Pennsylvania, a state that has slowly been shifting toward the Republicans for the last twenty years. Or, at least part of the state has been shifting red.
AT's political correspondent Rich Baehr believes at least part of the problem is that voters are lying to pollsters. There might be something to that, except most good pollsters throw trick questions into their surveys that are designed to reveal the liars. Baehr thinks that Americans have become very sophisticated when it comes to polls and have learned how to give answers they think makes them look good, rather than what they truly believe.
It's an interesting theory that we will test tomorrow, as we will see how the state and national polls line up in the aftermath. It could be so close that the difference between the polls and reality might be explained by enthusiasm, or turnout, or a superior ground game by one side or the other. But if it's not very close, pollsters will have to re-examine their methods and models to determine where they went so wrong.